Uncertainty Matters in Dynamic Gaussian Splatting for Monocular 4D Reconstruction

2025-10-14

AI Summary

Monocular dynamic 3D reconstruction suffers from severe under-constrained optimization due to occlusions and extreme novel viewpoints, leading to motion drift and degraded rendering quality in conventional dynamic Gaussian splatting methods. To address this, we propose an uncertainty-aware 4D Gaussian splatting framework. First, we introduce a time-adaptive per-Gaussian uncertainty estimation module that quantifies the observational reliability of each Gaussian primitive. Second, we design an uncertainty-weighted spatiotemporal graph optimization mechanism that propagates motion cues exclusively through high-confidence anchor Gaussians, thereby suppressing error propagation. Our method significantly improves geometric stability on both synthetic and real-world monocular dynamic video sequences. It effectively mitigates occlusion-induced motion drift and enhances novel view synthesis quality under extreme viewpoints. The source code and datasets will be made publicly available.

Technology Category

• Computer Vision: Motion & Tracking
• Reasoning under Uncertainty: Uncertainty Representations
• Machine Learning: Calibration & Uncertainty Quantification

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Security and Privacy: Data transparency and provenance
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

Reconstructing dynamic 3D scenes from monocular input is fundamentally under-constrained, with ambiguities arising from occlusion and extreme novel views. While dynamic Gaussian Splatting offers an efficient representation, vanilla models optimize all Gaussian primitives uniformly, ignoring whether they are well or poorly observed. This limitation leads to motion drifts under occlusion and degraded synthesis when extrapolating to unseen views. We argue that uncertainty matters: Gaussians with recurring observations across views and time act as reliable anchors to guide motion, whereas those with limited visibility are treated as less reliable. To this end, we introduce USplat4D, a novel Uncertainty-aware dynamic Gaussian Splatting framework that propagates reliable motion cues to enhance 4D reconstruction. Our key insight is to estimate time-varying per-Gaussian uncertainty and leverages it to construct a spatio-temporal graph for uncertainty-aware optimization. Experiments on diverse real and synthetic datasets show that explicitly modeling uncertainty consistently improves dynamic Gaussian Splatting models, yielding more stable geometry under occlusion and high-quality synthesis at extreme viewpoints.

Problem

• Addresses motion drift under occlusion in dynamic 3D reconstruction
• Improves synthesis quality for unseen and extreme viewpoints
• Models per-Gaussian uncertainty to guide reliable motion propagation

Innovation

• Models per-Gaussian uncertainty for dynamic scenes
• Constructs spatio-temporal graph for optimization
• Propagates reliable motion cues to enhance reconstruction